1/*-
2 * Copyright (c) 1991, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Kenneth Almquist.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 4. Neither the name of the University nor the names of its contributors
17 *    may be used to endorse or promote products derived from this software
18 *    without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33#ifndef lint
34#if 0
35static char sccsid[] = "@(#)parser.c	8.7 (Berkeley) 5/16/95";
36#endif
37#endif /* not lint */
38#include <sys/cdefs.h>
39__FBSDID("$FreeBSD$");
40
41#include <stdlib.h>
42#include <unistd.h>
43#include <stdio.h>
44
45#include "shell.h"
46#include "parser.h"
47#include "nodes.h"
48#include "expand.h"	/* defines rmescapes() */
49#include "syntax.h"
50#include "options.h"
51#include "input.h"
52#include "output.h"
53#include "var.h"
54#include "error.h"
55#include "memalloc.h"
56#include "mystring.h"
57#include "alias.h"
58#include "show.h"
59#include "eval.h"
60#include "exec.h"	/* to check for special builtins */
61#ifndef NO_HISTORY
62#include "myhistedit.h"
63#endif
64
65/*
66 * Shell command parser.
67 */
68
69#define	EOFMARKLEN	79
70#define	PROMPTLEN	128
71
72/* values of checkkwd variable */
73#define CHKALIAS	0x1
74#define CHKKWD		0x2
75#define CHKNL		0x4
76
77/* values returned by readtoken */
78#include "token.h"
79
80
81
82struct heredoc {
83	struct heredoc *next;	/* next here document in list */
84	union node *here;		/* redirection node */
85	char *eofmark;		/* string indicating end of input */
86	int striptabs;		/* if set, strip leading tabs */
87};
88
89struct parser_temp {
90	struct parser_temp *next;
91	void *data;
92};
93
94
95static struct heredoc *heredoclist;	/* list of here documents to read */
96static int doprompt;		/* if set, prompt the user */
97static int needprompt;		/* true if interactive and at start of line */
98static int lasttoken;		/* last token read */
99static int tokpushback;		/* last token pushed back */
100static char *wordtext;		/* text of last word returned by readtoken */
101static int checkkwd;
102static struct nodelist *backquotelist;
103static union node *redirnode;
104static struct heredoc *heredoc;
105static int quoteflag;		/* set if (part of) last token was quoted */
106static int startlinno;		/* line # where last token started */
107static int funclinno;		/* line # where the current function started */
108static struct parser_temp *parser_temp;
109
110
111static union node *list(int);
112static union node *andor(void);
113static union node *pipeline(void);
114static union node *command(void);
115static union node *simplecmd(union node **, union node *);
116static union node *makename(void);
117static union node *makebinary(int type, union node *n1, union node *n2);
118static void parsefname(void);
119static void parseheredoc(void);
120static int peektoken(void);
121static int readtoken(void);
122static int xxreadtoken(void);
123static int readtoken1(int, const char *, const char *, int);
124static int noexpand(char *);
125static void consumetoken(int);
126static void synexpect(int) __dead2;
127static void synerror(const char *) __dead2;
128static void setprompt(int);
129
130
131static void *
132parser_temp_alloc(size_t len)
133{
134	struct parser_temp *t;
135
136	INTOFF;
137	t = ckmalloc(sizeof(*t));
138	t->data = NULL;
139	t->next = parser_temp;
140	parser_temp = t;
141	t->data = ckmalloc(len);
142	INTON;
143	return t->data;
144}
145
146
147static void *
148parser_temp_realloc(void *ptr, size_t len)
149{
150	struct parser_temp *t;
151
152	INTOFF;
153	t = parser_temp;
154	if (ptr != t->data)
155		error("bug: parser_temp_realloc misused");
156	t->data = ckrealloc(t->data, len);
157	INTON;
158	return t->data;
159}
160
161
162static void
163parser_temp_free_upto(void *ptr)
164{
165	struct parser_temp *t;
166	int done = 0;
167
168	INTOFF;
169	while (parser_temp != NULL && !done) {
170		t = parser_temp;
171		parser_temp = t->next;
172		done = t->data == ptr;
173		ckfree(t->data);
174		ckfree(t);
175	}
176	INTON;
177	if (!done)
178		error("bug: parser_temp_free_upto misused");
179}
180
181
182static void
183parser_temp_free_all(void)
184{
185	struct parser_temp *t;
186
187	INTOFF;
188	while (parser_temp != NULL) {
189		t = parser_temp;
190		parser_temp = t->next;
191		ckfree(t->data);
192		ckfree(t);
193	}
194	INTON;
195}
196
197
198/*
199 * Read and parse a command.  Returns NEOF on end of file.  (NULL is a
200 * valid parse tree indicating a blank line.)
201 */
202
203union node *
204parsecmd(int interact)
205{
206	int t;
207
208	/* This assumes the parser is not re-entered,
209	 * which could happen if we add command substitution on PS1/PS2.
210	 */
211	parser_temp_free_all();
212	heredoclist = NULL;
213
214	tokpushback = 0;
215	checkkwd = 0;
216	doprompt = interact;
217	if (doprompt)
218		setprompt(1);
219	else
220		setprompt(0);
221	needprompt = 0;
222	t = readtoken();
223	if (t == TEOF)
224		return NEOF;
225	if (t == TNL)
226		return NULL;
227	tokpushback++;
228	return list(1);
229}
230
231
232static union node *
233list(int nlflag)
234{
235	union node *ntop, *n1, *n2, *n3;
236	int tok;
237
238	checkkwd = CHKNL | CHKKWD | CHKALIAS;
239	if (!nlflag && tokendlist[peektoken()])
240		return NULL;
241	ntop = n1 = NULL;
242	for (;;) {
243		n2 = andor();
244		tok = readtoken();
245		if (tok == TBACKGND) {
246			if (n2 != NULL && n2->type == NPIPE) {
247				n2->npipe.backgnd = 1;
248			} else if (n2 != NULL && n2->type == NREDIR) {
249				n2->type = NBACKGND;
250			} else {
251				n3 = (union node *)stalloc(sizeof (struct nredir));
252				n3->type = NBACKGND;
253				n3->nredir.n = n2;
254				n3->nredir.redirect = NULL;
255				n2 = n3;
256			}
257		}
258		if (ntop == NULL)
259			ntop = n2;
260		else if (n1 == NULL) {
261			n1 = makebinary(NSEMI, ntop, n2);
262			ntop = n1;
263		}
264		else {
265			n3 = makebinary(NSEMI, n1->nbinary.ch2, n2);
266			n1->nbinary.ch2 = n3;
267			n1 = n3;
268		}
269		switch (tok) {
270		case TBACKGND:
271		case TSEMI:
272			tok = readtoken();
273			/* FALLTHROUGH */
274		case TNL:
275			if (tok == TNL) {
276				parseheredoc();
277				if (nlflag)
278					return ntop;
279			} else if (tok == TEOF && nlflag) {
280				parseheredoc();
281				return ntop;
282			} else {
283				tokpushback++;
284			}
285			checkkwd = CHKNL | CHKKWD | CHKALIAS;
286			if (!nlflag && tokendlist[peektoken()])
287				return ntop;
288			break;
289		case TEOF:
290			if (heredoclist)
291				parseheredoc();
292			else
293				pungetc();		/* push back EOF on input */
294			return ntop;
295		default:
296			if (nlflag)
297				synexpect(-1);
298			tokpushback++;
299			return ntop;
300		}
301	}
302}
303
304
305
306static union node *
307andor(void)
308{
309	union node *n;
310	int t;
311
312	n = pipeline();
313	for (;;) {
314		if ((t = readtoken()) == TAND) {
315			t = NAND;
316		} else if (t == TOR) {
317			t = NOR;
318		} else {
319			tokpushback++;
320			return n;
321		}
322		n = makebinary(t, n, pipeline());
323	}
324}
325
326
327
328static union node *
329pipeline(void)
330{
331	union node *n1, *n2, *pipenode;
332	struct nodelist *lp, *prev;
333	int negate, t;
334
335	negate = 0;
336	checkkwd = CHKNL | CHKKWD | CHKALIAS;
337	TRACE(("pipeline: entered\n"));
338	while (readtoken() == TNOT)
339		negate = !negate;
340	tokpushback++;
341	n1 = command();
342	if (readtoken() == TPIPE) {
343		pipenode = (union node *)stalloc(sizeof (struct npipe));
344		pipenode->type = NPIPE;
345		pipenode->npipe.backgnd = 0;
346		lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
347		pipenode->npipe.cmdlist = lp;
348		lp->n = n1;
349		do {
350			prev = lp;
351			lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
352			checkkwd = CHKNL | CHKKWD | CHKALIAS;
353			t = readtoken();
354			tokpushback++;
355			if (t == TNOT)
356				lp->n = pipeline();
357			else
358				lp->n = command();
359			prev->next = lp;
360		} while (readtoken() == TPIPE);
361		lp->next = NULL;
362		n1 = pipenode;
363	}
364	tokpushback++;
365	if (negate) {
366		n2 = (union node *)stalloc(sizeof (struct nnot));
367		n2->type = NNOT;
368		n2->nnot.com = n1;
369		return n2;
370	} else
371		return n1;
372}
373
374
375
376static union node *
377command(void)
378{
379	union node *n1, *n2;
380	union node *ap, **app;
381	union node *cp, **cpp;
382	union node *redir, **rpp;
383	int t;
384	int is_subshell;
385
386	checkkwd = CHKNL | CHKKWD | CHKALIAS;
387	is_subshell = 0;
388	redir = NULL;
389	n1 = NULL;
390	rpp = &redir;
391
392	/* Check for redirection which may precede command */
393	while (readtoken() == TREDIR) {
394		*rpp = n2 = redirnode;
395		rpp = &n2->nfile.next;
396		parsefname();
397	}
398	tokpushback++;
399
400	switch (readtoken()) {
401	case TIF:
402		n1 = (union node *)stalloc(sizeof (struct nif));
403		n1->type = NIF;
404		if ((n1->nif.test = list(0)) == NULL)
405			synexpect(-1);
406		consumetoken(TTHEN);
407		n1->nif.ifpart = list(0);
408		n2 = n1;
409		while (readtoken() == TELIF) {
410			n2->nif.elsepart = (union node *)stalloc(sizeof (struct nif));
411			n2 = n2->nif.elsepart;
412			n2->type = NIF;
413			if ((n2->nif.test = list(0)) == NULL)
414				synexpect(-1);
415			consumetoken(TTHEN);
416			n2->nif.ifpart = list(0);
417		}
418		if (lasttoken == TELSE)
419			n2->nif.elsepart = list(0);
420		else {
421			n2->nif.elsepart = NULL;
422			tokpushback++;
423		}
424		consumetoken(TFI);
425		checkkwd = CHKKWD | CHKALIAS;
426		break;
427	case TWHILE:
428	case TUNTIL:
429		t = lasttoken;
430		if ((n1 = list(0)) == NULL)
431			synexpect(-1);
432		consumetoken(TDO);
433		n1 = makebinary((t == TWHILE)? NWHILE : NUNTIL, n1, list(0));
434		consumetoken(TDONE);
435		checkkwd = CHKKWD | CHKALIAS;
436		break;
437	case TFOR:
438		if (readtoken() != TWORD || quoteflag || ! goodname(wordtext))
439			synerror("Bad for loop variable");
440		n1 = (union node *)stalloc(sizeof (struct nfor));
441		n1->type = NFOR;
442		n1->nfor.var = wordtext;
443		while (readtoken() == TNL)
444			;
445		if (lasttoken == TWORD && ! quoteflag && equal(wordtext, "in")) {
446			app = &ap;
447			while (readtoken() == TWORD) {
448				n2 = makename();
449				*app = n2;
450				app = &n2->narg.next;
451			}
452			*app = NULL;
453			n1->nfor.args = ap;
454			if (lasttoken != TNL && lasttoken != TSEMI)
455				synexpect(-1);
456		} else {
457			static char argvars[5] = {
458				CTLVAR, VSNORMAL|VSQUOTE, '@', '=', '\0'
459			};
460			n2 = (union node *)stalloc(sizeof (struct narg));
461			n2->type = NARG;
462			n2->narg.text = argvars;
463			n2->narg.backquote = NULL;
464			n2->narg.next = NULL;
465			n1->nfor.args = n2;
466			/*
467			 * Newline or semicolon here is optional (but note
468			 * that the original Bourne shell only allowed NL).
469			 */
470			if (lasttoken != TNL && lasttoken != TSEMI)
471				tokpushback++;
472		}
473		checkkwd = CHKNL | CHKKWD | CHKALIAS;
474		if ((t = readtoken()) == TDO)
475			t = TDONE;
476		else if (t == TBEGIN)
477			t = TEND;
478		else
479			synexpect(-1);
480		n1->nfor.body = list(0);
481		consumetoken(t);
482		checkkwd = CHKKWD | CHKALIAS;
483		break;
484	case TCASE:
485		n1 = (union node *)stalloc(sizeof (struct ncase));
486		n1->type = NCASE;
487		consumetoken(TWORD);
488		n1->ncase.expr = makename();
489		while (readtoken() == TNL);
490		if (lasttoken != TWORD || ! equal(wordtext, "in"))
491			synerror("expecting \"in\"");
492		cpp = &n1->ncase.cases;
493		checkkwd = CHKNL | CHKKWD, readtoken();
494		while (lasttoken != TESAC) {
495			*cpp = cp = (union node *)stalloc(sizeof (struct nclist));
496			cp->type = NCLIST;
497			app = &cp->nclist.pattern;
498			if (lasttoken == TLP)
499				readtoken();
500			for (;;) {
501				*app = ap = makename();
502				checkkwd = CHKNL | CHKKWD;
503				if (readtoken() != TPIPE)
504					break;
505				app = &ap->narg.next;
506				readtoken();
507			}
508			ap->narg.next = NULL;
509			if (lasttoken != TRP)
510				synexpect(TRP);
511			cp->nclist.body = list(0);
512
513			checkkwd = CHKNL | CHKKWD | CHKALIAS;
514			if ((t = readtoken()) != TESAC) {
515				if (t == TENDCASE)
516					;
517				else if (t == TFALLTHRU)
518					cp->type = NCLISTFALLTHRU;
519				else
520					synexpect(TENDCASE);
521				checkkwd = CHKNL | CHKKWD, readtoken();
522			}
523			cpp = &cp->nclist.next;
524		}
525		*cpp = NULL;
526		checkkwd = CHKKWD | CHKALIAS;
527		break;
528	case TLP:
529		n1 = (union node *)stalloc(sizeof (struct nredir));
530		n1->type = NSUBSHELL;
531		n1->nredir.n = list(0);
532		n1->nredir.redirect = NULL;
533		consumetoken(TRP);
534		checkkwd = CHKKWD | CHKALIAS;
535		is_subshell = 1;
536		break;
537	case TBEGIN:
538		n1 = list(0);
539		consumetoken(TEND);
540		checkkwd = CHKKWD | CHKALIAS;
541		break;
542	/* A simple command must have at least one redirection or word. */
543	case TBACKGND:
544	case TSEMI:
545	case TAND:
546	case TOR:
547	case TPIPE:
548	case TENDCASE:
549	case TFALLTHRU:
550	case TEOF:
551	case TNL:
552	case TRP:
553		if (!redir)
554			synexpect(-1);
555	case TWORD:
556		tokpushback++;
557		n1 = simplecmd(rpp, redir);
558		return n1;
559	default:
560		synexpect(-1);
561	}
562
563	/* Now check for redirection which may follow command */
564	while (readtoken() == TREDIR) {
565		*rpp = n2 = redirnode;
566		rpp = &n2->nfile.next;
567		parsefname();
568	}
569	tokpushback++;
570	*rpp = NULL;
571	if (redir) {
572		if (!is_subshell) {
573			n2 = (union node *)stalloc(sizeof (struct nredir));
574			n2->type = NREDIR;
575			n2->nredir.n = n1;
576			n1 = n2;
577		}
578		n1->nredir.redirect = redir;
579	}
580
581	return n1;
582}
583
584
585static union node *
586simplecmd(union node **rpp, union node *redir)
587{
588	union node *args, **app;
589	union node **orig_rpp = rpp;
590	union node *n = NULL;
591	int special;
592	int savecheckkwd;
593
594	/* If we don't have any redirections already, then we must reset */
595	/* rpp to be the address of the local redir variable.  */
596	if (redir == 0)
597		rpp = &redir;
598
599	args = NULL;
600	app = &args;
601	/*
602	 * We save the incoming value, because we need this for shell
603	 * functions.  There can not be a redirect or an argument between
604	 * the function name and the open parenthesis.
605	 */
606	orig_rpp = rpp;
607
608	savecheckkwd = CHKALIAS;
609
610	for (;;) {
611		checkkwd = savecheckkwd;
612		if (readtoken() == TWORD) {
613			n = makename();
614			*app = n;
615			app = &n->narg.next;
616			if (savecheckkwd != 0 && !isassignment(wordtext))
617				savecheckkwd = 0;
618		} else if (lasttoken == TREDIR) {
619			*rpp = n = redirnode;
620			rpp = &n->nfile.next;
621			parsefname();	/* read name of redirection file */
622		} else if (lasttoken == TLP && app == &args->narg.next
623					    && rpp == orig_rpp) {
624			/* We have a function */
625			consumetoken(TRP);
626			funclinno = plinno;
627			/*
628			 * - Require plain text.
629			 * - Functions with '/' cannot be called.
630			 * - Reject name=().
631			 * - Reject ksh extended glob patterns.
632			 */
633			if (!noexpand(n->narg.text) || quoteflag ||
634			    strchr(n->narg.text, '/') ||
635			    strchr("!%*+-=?@}~",
636				n->narg.text[strlen(n->narg.text) - 1]))
637				synerror("Bad function name");
638			rmescapes(n->narg.text);
639			if (find_builtin(n->narg.text, &special) >= 0 &&
640			    special)
641				synerror("Cannot override a special builtin with a function");
642			n->type = NDEFUN;
643			n->narg.next = command();
644			funclinno = 0;
645			return n;
646		} else {
647			tokpushback++;
648			break;
649		}
650	}
651	*app = NULL;
652	*rpp = NULL;
653	n = (union node *)stalloc(sizeof (struct ncmd));
654	n->type = NCMD;
655	n->ncmd.args = args;
656	n->ncmd.redirect = redir;
657	return n;
658}
659
660static union node *
661makename(void)
662{
663	union node *n;
664
665	n = (union node *)stalloc(sizeof (struct narg));
666	n->type = NARG;
667	n->narg.next = NULL;
668	n->narg.text = wordtext;
669	n->narg.backquote = backquotelist;
670	return n;
671}
672
673static union node *
674makebinary(int type, union node *n1, union node *n2)
675{
676	union node *n;
677
678	n = (union node *)stalloc(sizeof (struct nbinary));
679	n->type = type;
680	n->nbinary.ch1 = n1;
681	n->nbinary.ch2 = n2;
682	return (n);
683}
684
685void
686fixredir(union node *n, const char *text, int err)
687{
688	TRACE(("Fix redir %s %d\n", text, err));
689	if (!err)
690		n->ndup.vname = NULL;
691
692	if (is_digit(text[0]) && text[1] == '\0')
693		n->ndup.dupfd = digit_val(text[0]);
694	else if (text[0] == '-' && text[1] == '\0')
695		n->ndup.dupfd = -1;
696	else {
697
698		if (err)
699			synerror("Bad fd number");
700		else
701			n->ndup.vname = makename();
702	}
703}
704
705
706static void
707parsefname(void)
708{
709	union node *n = redirnode;
710
711	consumetoken(TWORD);
712	if (n->type == NHERE) {
713		struct heredoc *here = heredoc;
714		struct heredoc *p;
715		int i;
716
717		if (quoteflag == 0)
718			n->type = NXHERE;
719		TRACE(("Here document %d\n", n->type));
720		if (here->striptabs) {
721			while (*wordtext == '\t')
722				wordtext++;
723		}
724		if (! noexpand(wordtext) || (i = strlen(wordtext)) == 0 || i > EOFMARKLEN)
725			synerror("Illegal eof marker for << redirection");
726		rmescapes(wordtext);
727		here->eofmark = wordtext;
728		here->next = NULL;
729		if (heredoclist == NULL)
730			heredoclist = here;
731		else {
732			for (p = heredoclist ; p->next ; p = p->next);
733			p->next = here;
734		}
735	} else if (n->type == NTOFD || n->type == NFROMFD) {
736		fixredir(n, wordtext, 0);
737	} else {
738		n->nfile.fname = makename();
739	}
740}
741
742
743/*
744 * Input any here documents.
745 */
746
747static void
748parseheredoc(void)
749{
750	struct heredoc *here;
751	union node *n;
752
753	while (heredoclist) {
754		here = heredoclist;
755		heredoclist = here->next;
756		if (needprompt) {
757			setprompt(2);
758			needprompt = 0;
759		}
760		readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX,
761				here->eofmark, here->striptabs);
762		n = makename();
763		here->here->nhere.doc = n;
764	}
765}
766
767static int
768peektoken(void)
769{
770	int t;
771
772	t = readtoken();
773	tokpushback++;
774	return (t);
775}
776
777static int
778readtoken(void)
779{
780	int t;
781	struct alias *ap;
782#ifdef DEBUG
783	int alreadyseen = tokpushback;
784#endif
785
786	top:
787	t = xxreadtoken();
788
789	/*
790	 * eat newlines
791	 */
792	if (checkkwd & CHKNL) {
793		while (t == TNL) {
794			parseheredoc();
795			t = xxreadtoken();
796		}
797	}
798
799	/*
800	 * check for keywords and aliases
801	 */
802	if (t == TWORD && !quoteflag)
803	{
804		const char * const *pp;
805
806		if (checkkwd & CHKKWD)
807			for (pp = parsekwd; *pp; pp++) {
808				if (**pp == *wordtext && equal(*pp, wordtext))
809				{
810					lasttoken = t = pp - parsekwd + KWDOFFSET;
811					TRACE(("keyword %s recognized\n", tokname[t]));
812					goto out;
813				}
814			}
815		if (checkkwd & CHKALIAS &&
816		    (ap = lookupalias(wordtext, 1)) != NULL) {
817			pushstring(ap->val, strlen(ap->val), ap);
818			goto top;
819		}
820	}
821out:
822	if (t != TNOT)
823		checkkwd = 0;
824
825#ifdef DEBUG
826	if (!alreadyseen)
827	    TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
828	else
829	    TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
830#endif
831	return (t);
832}
833
834
835/*
836 * Read the next input token.
837 * If the token is a word, we set backquotelist to the list of cmds in
838 *	backquotes.  We set quoteflag to true if any part of the word was
839 *	quoted.
840 * If the token is TREDIR, then we set redirnode to a structure containing
841 *	the redirection.
842 * In all cases, the variable startlinno is set to the number of the line
843 *	on which the token starts.
844 *
845 * [Change comment:  here documents and internal procedures]
846 * [Readtoken shouldn't have any arguments.  Perhaps we should make the
847 *  word parsing code into a separate routine.  In this case, readtoken
848 *  doesn't need to have any internal procedures, but parseword does.
849 *  We could also make parseoperator in essence the main routine, and
850 *  have parseword (readtoken1?) handle both words and redirection.]
851 */
852
853#define RETURN(token)	return lasttoken = token
854
855static int
856xxreadtoken(void)
857{
858	int c;
859
860	if (tokpushback) {
861		tokpushback = 0;
862		return lasttoken;
863	}
864	if (needprompt) {
865		setprompt(2);
866		needprompt = 0;
867	}
868	startlinno = plinno;
869	for (;;) {	/* until token or start of word found */
870		c = pgetc_macro();
871		switch (c) {
872		case ' ': case '\t':
873			continue;
874		case '#':
875			while ((c = pgetc()) != '\n' && c != PEOF);
876			pungetc();
877			continue;
878		case '\\':
879			if (pgetc() == '\n') {
880				startlinno = ++plinno;
881				if (doprompt)
882					setprompt(2);
883				else
884					setprompt(0);
885				continue;
886			}
887			pungetc();
888			goto breakloop;
889		case '\n':
890			plinno++;
891			needprompt = doprompt;
892			RETURN(TNL);
893		case PEOF:
894			RETURN(TEOF);
895		case '&':
896			if (pgetc() == '&')
897				RETURN(TAND);
898			pungetc();
899			RETURN(TBACKGND);
900		case '|':
901			if (pgetc() == '|')
902				RETURN(TOR);
903			pungetc();
904			RETURN(TPIPE);
905		case ';':
906			c = pgetc();
907			if (c == ';')
908				RETURN(TENDCASE);
909			else if (c == '&')
910				RETURN(TFALLTHRU);
911			pungetc();
912			RETURN(TSEMI);
913		case '(':
914			RETURN(TLP);
915		case ')':
916			RETURN(TRP);
917		default:
918			goto breakloop;
919		}
920	}
921breakloop:
922	return readtoken1(c, BASESYNTAX, (char *)NULL, 0);
923#undef RETURN
924}
925
926
927#define MAXNEST_static 8
928struct tokenstate
929{
930	const char *syntax; /* *SYNTAX */
931	int parenlevel; /* levels of parentheses in arithmetic */
932	enum tokenstate_category
933	{
934		TSTATE_TOP,
935		TSTATE_VAR_OLD, /* ${var+-=?}, inherits dquotes */
936		TSTATE_VAR_NEW, /* other ${var...}, own dquote state */
937		TSTATE_ARITH
938	} category;
939};
940
941
942/*
943 * Called to parse command substitutions.
944 */
945
946static char *
947parsebackq(char *out, struct nodelist **pbqlist,
948		int oldstyle, int dblquote, int quoted)
949{
950	struct nodelist **nlpp;
951	union node *n;
952	char *volatile str;
953	struct jmploc jmploc;
954	struct jmploc *const savehandler = handler;
955	size_t savelen;
956	int saveprompt;
957	const int bq_startlinno = plinno;
958	char *volatile ostr = NULL;
959	struct parsefile *const savetopfile = getcurrentfile();
960	struct heredoc *const saveheredoclist = heredoclist;
961	struct heredoc *here;
962
963	str = NULL;
964	if (setjmp(jmploc.loc)) {
965		popfilesupto(savetopfile);
966		if (str)
967			ckfree(str);
968		if (ostr)
969			ckfree(ostr);
970		heredoclist = saveheredoclist;
971		handler = savehandler;
972		if (exception == EXERROR) {
973			startlinno = bq_startlinno;
974			synerror("Error in command substitution");
975		}
976		longjmp(handler->loc, 1);
977	}
978	INTOFF;
979	savelen = out - stackblock();
980	if (savelen > 0) {
981		str = ckmalloc(savelen);
982		memcpy(str, stackblock(), savelen);
983	}
984	handler = &jmploc;
985	heredoclist = NULL;
986	INTON;
987        if (oldstyle) {
988                /* We must read until the closing backquote, giving special
989                   treatment to some slashes, and then push the string and
990                   reread it as input, interpreting it normally.  */
991                char *oout;
992                int c;
993                int olen;
994
995
996                STARTSTACKSTR(oout);
997		for (;;) {
998			if (needprompt) {
999				setprompt(2);
1000				needprompt = 0;
1001			}
1002			CHECKSTRSPACE(2, oout);
1003			switch (c = pgetc()) {
1004			case '`':
1005				goto done;
1006
1007			case '\\':
1008                                if ((c = pgetc()) == '\n') {
1009					plinno++;
1010					if (doprompt)
1011						setprompt(2);
1012					else
1013						setprompt(0);
1014					/*
1015					 * If eating a newline, avoid putting
1016					 * the newline into the new character
1017					 * stream (via the USTPUTC after the
1018					 * switch).
1019					 */
1020					continue;
1021				}
1022                                if (c != '\\' && c != '`' && c != '$'
1023                                    && (!dblquote || c != '"'))
1024                                        USTPUTC('\\', oout);
1025				break;
1026
1027			case '\n':
1028				plinno++;
1029				needprompt = doprompt;
1030				break;
1031
1032			case PEOF:
1033			        startlinno = plinno;
1034				synerror("EOF in backquote substitution");
1035 				break;
1036
1037			default:
1038				break;
1039			}
1040			USTPUTC(c, oout);
1041                }
1042done:
1043                USTPUTC('\0', oout);
1044                olen = oout - stackblock();
1045		INTOFF;
1046		ostr = ckmalloc(olen);
1047		memcpy(ostr, stackblock(), olen);
1048		setinputstring(ostr, 1);
1049		INTON;
1050        }
1051	nlpp = pbqlist;
1052	while (*nlpp)
1053		nlpp = &(*nlpp)->next;
1054	*nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist));
1055	(*nlpp)->next = NULL;
1056
1057	if (oldstyle) {
1058		saveprompt = doprompt;
1059		doprompt = 0;
1060	}
1061
1062	n = list(0);
1063
1064	if (oldstyle) {
1065		if (peektoken() != TEOF)
1066			synexpect(-1);
1067		doprompt = saveprompt;
1068	} else
1069		consumetoken(TRP);
1070
1071	(*nlpp)->n = n;
1072        if (oldstyle) {
1073		/*
1074		 * Start reading from old file again, ignoring any pushed back
1075		 * tokens left from the backquote parsing
1076		 */
1077                popfile();
1078		tokpushback = 0;
1079	}
1080	STARTSTACKSTR(out);
1081	CHECKSTRSPACE(savelen + 1, out);
1082	INTOFF;
1083	if (str) {
1084		memcpy(out, str, savelen);
1085		STADJUST(savelen, out);
1086		ckfree(str);
1087		str = NULL;
1088	}
1089	if (ostr) {
1090		ckfree(ostr);
1091		ostr = NULL;
1092	}
1093	here = saveheredoclist;
1094	if (here != NULL) {
1095		while (here->next != NULL)
1096			here = here->next;
1097		here->next = heredoclist;
1098		heredoclist = saveheredoclist;
1099	}
1100	handler = savehandler;
1101	INTON;
1102	if (quoted)
1103		USTPUTC(CTLBACKQ | CTLQUOTE, out);
1104	else
1105		USTPUTC(CTLBACKQ, out);
1106	return out;
1107}
1108
1109
1110/*
1111 * Called to parse a backslash escape sequence inside $'...'.
1112 * The backslash has already been read.
1113 */
1114static char *
1115readcstyleesc(char *out)
1116{
1117	int c, v, i, n;
1118
1119	c = pgetc();
1120	switch (c) {
1121	case '\0':
1122		synerror("Unterminated quoted string");
1123	case '\n':
1124		plinno++;
1125		if (doprompt)
1126			setprompt(2);
1127		else
1128			setprompt(0);
1129		return out;
1130	case '\\':
1131	case '\'':
1132	case '"':
1133		v = c;
1134		break;
1135	case 'a': v = '\a'; break;
1136	case 'b': v = '\b'; break;
1137	case 'e': v = '\033'; break;
1138	case 'f': v = '\f'; break;
1139	case 'n': v = '\n'; break;
1140	case 'r': v = '\r'; break;
1141	case 't': v = '\t'; break;
1142	case 'v': v = '\v'; break;
1143	case 'x':
1144		  v = 0;
1145		  for (;;) {
1146			  c = pgetc();
1147			  if (c >= '0' && c <= '9')
1148				  v = (v << 4) + c - '0';
1149			  else if (c >= 'A' && c <= 'F')
1150				  v = (v << 4) + c - 'A' + 10;
1151			  else if (c >= 'a' && c <= 'f')
1152				  v = (v << 4) + c - 'a' + 10;
1153			  else
1154				  break;
1155		  }
1156		  pungetc();
1157		  break;
1158	case '0': case '1': case '2': case '3':
1159	case '4': case '5': case '6': case '7':
1160		  v = c - '0';
1161		  c = pgetc();
1162		  if (c >= '0' && c <= '7') {
1163			  v <<= 3;
1164			  v += c - '0';
1165			  c = pgetc();
1166			  if (c >= '0' && c <= '7') {
1167				  v <<= 3;
1168				  v += c - '0';
1169			  } else
1170				  pungetc();
1171		  } else
1172			  pungetc();
1173		  break;
1174	case 'c':
1175		  c = pgetc();
1176		  if (c < 0x3f || c > 0x7a || c == 0x60)
1177			  synerror("Bad escape sequence");
1178		  if (c == '\\' && pgetc() != '\\')
1179			  synerror("Bad escape sequence");
1180		  if (c == '?')
1181			  v = 127;
1182		  else
1183			  v = c & 0x1f;
1184		  break;
1185	case 'u':
1186	case 'U':
1187		  n = c == 'U' ? 8 : 4;
1188		  v = 0;
1189		  for (i = 0; i < n; i++) {
1190			  c = pgetc();
1191			  if (c >= '0' && c <= '9')
1192				  v = (v << 4) + c - '0';
1193			  else if (c >= 'A' && c <= 'F')
1194				  v = (v << 4) + c - 'A' + 10;
1195			  else if (c >= 'a' && c <= 'f')
1196				  v = (v << 4) + c - 'a' + 10;
1197			  else
1198				  synerror("Bad escape sequence");
1199		  }
1200		  if (v == 0 || (v >= 0xd800 && v <= 0xdfff))
1201			  synerror("Bad escape sequence");
1202		  /* We really need iconv here. */
1203		  if (initial_localeisutf8 && v > 127) {
1204			  CHECKSTRSPACE(4, out);
1205			  /*
1206			   * We cannot use wctomb() as the locale may have
1207			   * changed.
1208			   */
1209			  if (v <= 0x7ff) {
1210				  USTPUTC(0xc0 | v >> 6, out);
1211				  USTPUTC(0x80 | (v & 0x3f), out);
1212				  return out;
1213			  } else if (v <= 0xffff) {
1214				  USTPUTC(0xe0 | v >> 12, out);
1215				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1216				  USTPUTC(0x80 | (v & 0x3f), out);
1217				  return out;
1218			  } else if (v <= 0x10ffff) {
1219				  USTPUTC(0xf0 | v >> 18, out);
1220				  USTPUTC(0x80 | ((v >> 12) & 0x3f), out);
1221				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1222				  USTPUTC(0x80 | (v & 0x3f), out);
1223				  return out;
1224			  }
1225		  }
1226		  if (v > 127)
1227			  v = '?';
1228		  break;
1229	default:
1230		  synerror("Bad escape sequence");
1231	}
1232	v = (char)v;
1233	/*
1234	 * We can't handle NUL bytes.
1235	 * POSIX says we should skip till the closing quote.
1236	 */
1237	if (v == '\0') {
1238		while ((c = pgetc()) != '\'') {
1239			if (c == '\\')
1240				c = pgetc();
1241			if (c == PEOF)
1242				synerror("Unterminated quoted string");
1243		}
1244		pungetc();
1245		return out;
1246	}
1247	if (SQSYNTAX[v] == CCTL)
1248		USTPUTC(CTLESC, out);
1249	USTPUTC(v, out);
1250	return out;
1251}
1252
1253
1254/*
1255 * If eofmark is NULL, read a word or a redirection symbol.  If eofmark
1256 * is not NULL, read a here document.  In the latter case, eofmark is the
1257 * word which marks the end of the document and striptabs is true if
1258 * leading tabs should be stripped from the document.  The argument firstc
1259 * is the first character of the input token or document.
1260 *
1261 * Because C does not have internal subroutines, I have simulated them
1262 * using goto's to implement the subroutine linkage.  The following macros
1263 * will run code that appears at the end of readtoken1.
1264 */
1265
1266#define CHECKEND()	{goto checkend; checkend_return:;}
1267#define PARSEREDIR()	{goto parseredir; parseredir_return:;}
1268#define PARSESUB()	{goto parsesub; parsesub_return:;}
1269#define	PARSEARITH()	{goto parsearith; parsearith_return:;}
1270
1271static int
1272readtoken1(int firstc, char const *initialsyntax, const char *eofmark,
1273    int striptabs)
1274{
1275	int c = firstc;
1276	char *out;
1277	int len;
1278	char line[EOFMARKLEN + 1];
1279	struct nodelist *bqlist;
1280	int quotef;
1281	int newvarnest;
1282	int level;
1283	int synentry;
1284	struct tokenstate state_static[MAXNEST_static];
1285	int maxnest = MAXNEST_static;
1286	struct tokenstate *state = state_static;
1287	int sqiscstyle = 0;
1288
1289	startlinno = plinno;
1290	quotef = 0;
1291	bqlist = NULL;
1292	newvarnest = 0;
1293	level = 0;
1294	state[level].syntax = initialsyntax;
1295	state[level].parenlevel = 0;
1296	state[level].category = TSTATE_TOP;
1297
1298	STARTSTACKSTR(out);
1299	loop: {	/* for each line, until end of word */
1300		CHECKEND();	/* set c to PEOF if at end of here document */
1301		for (;;) {	/* until end of line or end of word */
1302			CHECKSTRSPACE(4, out);	/* permit 4 calls to USTPUTC */
1303
1304			synentry = state[level].syntax[c];
1305
1306			switch(synentry) {
1307			case CNL:	/* '\n' */
1308				if (state[level].syntax == BASESYNTAX)
1309					goto endword;	/* exit outer loop */
1310				USTPUTC(c, out);
1311				plinno++;
1312				if (doprompt)
1313					setprompt(2);
1314				else
1315					setprompt(0);
1316				c = pgetc();
1317				goto loop;		/* continue outer loop */
1318			case CSBACK:
1319				if (sqiscstyle) {
1320					out = readcstyleesc(out);
1321					break;
1322				}
1323				/* FALLTHROUGH */
1324			case CWORD:
1325				USTPUTC(c, out);
1326				break;
1327			case CCTL:
1328				if (eofmark == NULL || initialsyntax != SQSYNTAX)
1329					USTPUTC(CTLESC, out);
1330				USTPUTC(c, out);
1331				break;
1332			case CBACK:	/* backslash */
1333				c = pgetc();
1334				if (c == PEOF) {
1335					USTPUTC('\\', out);
1336					pungetc();
1337				} else if (c == '\n') {
1338					plinno++;
1339					if (doprompt)
1340						setprompt(2);
1341					else
1342						setprompt(0);
1343				} else {
1344					if (state[level].syntax == DQSYNTAX &&
1345					    c != '\\' && c != '`' && c != '$' &&
1346					    (c != '"' || (eofmark != NULL &&
1347						newvarnest == 0)) &&
1348					    (c != '}' || state[level].category != TSTATE_VAR_OLD))
1349						USTPUTC('\\', out);
1350					if ((eofmark == NULL ||
1351					    newvarnest > 0) &&
1352					    state[level].syntax == BASESYNTAX)
1353						USTPUTC(CTLQUOTEMARK, out);
1354					if (SQSYNTAX[c] == CCTL)
1355						USTPUTC(CTLESC, out);
1356					USTPUTC(c, out);
1357					if ((eofmark == NULL ||
1358					    newvarnest > 0) &&
1359					    state[level].syntax == BASESYNTAX &&
1360					    state[level].category == TSTATE_VAR_OLD)
1361						USTPUTC(CTLQUOTEEND, out);
1362					quotef++;
1363				}
1364				break;
1365			case CSQUOTE:
1366				USTPUTC(CTLQUOTEMARK, out);
1367				state[level].syntax = SQSYNTAX;
1368				sqiscstyle = 0;
1369				break;
1370			case CDQUOTE:
1371				USTPUTC(CTLQUOTEMARK, out);
1372				state[level].syntax = DQSYNTAX;
1373				break;
1374			case CENDQUOTE:
1375				if (eofmark != NULL && newvarnest == 0)
1376					USTPUTC(c, out);
1377				else {
1378					if (state[level].category == TSTATE_VAR_OLD)
1379						USTPUTC(CTLQUOTEEND, out);
1380					state[level].syntax = BASESYNTAX;
1381					quotef++;
1382				}
1383				break;
1384			case CVAR:	/* '$' */
1385				PARSESUB();		/* parse substitution */
1386				break;
1387			case CENDVAR:	/* '}' */
1388				if (level > 0 &&
1389				    ((state[level].category == TSTATE_VAR_OLD &&
1390				      state[level].syntax ==
1391				      state[level - 1].syntax) ||
1392				    (state[level].category == TSTATE_VAR_NEW &&
1393				     state[level].syntax == BASESYNTAX))) {
1394					if (state[level].category == TSTATE_VAR_NEW)
1395						newvarnest--;
1396					level--;
1397					USTPUTC(CTLENDVAR, out);
1398				} else {
1399					USTPUTC(c, out);
1400				}
1401				break;
1402			case CLP:	/* '(' in arithmetic */
1403				state[level].parenlevel++;
1404				USTPUTC(c, out);
1405				break;
1406			case CRP:	/* ')' in arithmetic */
1407				if (state[level].parenlevel > 0) {
1408					USTPUTC(c, out);
1409					--state[level].parenlevel;
1410				} else {
1411					if (pgetc() == ')') {
1412						if (level > 0 &&
1413						    state[level].category == TSTATE_ARITH) {
1414							level--;
1415							USTPUTC(CTLENDARI, out);
1416						} else
1417							USTPUTC(')', out);
1418					} else {
1419						/*
1420						 * unbalanced parens
1421						 *  (don't 2nd guess - no error)
1422						 */
1423						pungetc();
1424						USTPUTC(')', out);
1425					}
1426				}
1427				break;
1428			case CBQUOTE:	/* '`' */
1429				out = parsebackq(out, &bqlist, 1,
1430				    state[level].syntax == DQSYNTAX &&
1431				    (eofmark == NULL || newvarnest > 0),
1432				    state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX);
1433				break;
1434			case CEOF:
1435				goto endword;		/* exit outer loop */
1436			case CIGN:
1437				break;
1438			default:
1439				if (level == 0)
1440					goto endword;	/* exit outer loop */
1441				USTPUTC(c, out);
1442			}
1443			c = pgetc_macro();
1444		}
1445	}
1446endword:
1447	if (state[level].syntax == ARISYNTAX)
1448		synerror("Missing '))'");
1449	if (state[level].syntax != BASESYNTAX && eofmark == NULL)
1450		synerror("Unterminated quoted string");
1451	if (state[level].category == TSTATE_VAR_OLD ||
1452	    state[level].category == TSTATE_VAR_NEW) {
1453		startlinno = plinno;
1454		synerror("Missing '}'");
1455	}
1456	if (state != state_static)
1457		parser_temp_free_upto(state);
1458	USTPUTC('\0', out);
1459	len = out - stackblock();
1460	out = stackblock();
1461	if (eofmark == NULL) {
1462		if ((c == '>' || c == '<')
1463		 && quotef == 0
1464		 && len <= 2
1465		 && (*out == '\0' || is_digit(*out))) {
1466			PARSEREDIR();
1467			return lasttoken = TREDIR;
1468		} else {
1469			pungetc();
1470		}
1471	}
1472	quoteflag = quotef;
1473	backquotelist = bqlist;
1474	grabstackblock(len);
1475	wordtext = out;
1476	return lasttoken = TWORD;
1477/* end of readtoken routine */
1478
1479
1480/*
1481 * Check to see whether we are at the end of the here document.  When this
1482 * is called, c is set to the first character of the next input line.  If
1483 * we are at the end of the here document, this routine sets the c to PEOF.
1484 */
1485
1486checkend: {
1487	if (eofmark) {
1488		if (striptabs) {
1489			while (c == '\t')
1490				c = pgetc();
1491		}
1492		if (c == *eofmark) {
1493			if (pfgets(line, sizeof line) != NULL) {
1494				const char *p, *q;
1495
1496				p = line;
1497				for (q = eofmark + 1 ; *q && *p == *q ; p++, q++);
1498				if ((*p == '\0' || *p == '\n') && *q == '\0') {
1499					c = PEOF;
1500					if (*p == '\n') {
1501						plinno++;
1502						needprompt = doprompt;
1503					}
1504				} else {
1505					pushstring(line, strlen(line), NULL);
1506				}
1507			}
1508		}
1509	}
1510	goto checkend_return;
1511}
1512
1513
1514/*
1515 * Parse a redirection operator.  The variable "out" points to a string
1516 * specifying the fd to be redirected.  The variable "c" contains the
1517 * first character of the redirection operator.
1518 */
1519
1520parseredir: {
1521	char fd = *out;
1522	union node *np;
1523
1524	np = (union node *)stalloc(sizeof (struct nfile));
1525	if (c == '>') {
1526		np->nfile.fd = 1;
1527		c = pgetc();
1528		if (c == '>')
1529			np->type = NAPPEND;
1530		else if (c == '&')
1531			np->type = NTOFD;
1532		else if (c == '|')
1533			np->type = NCLOBBER;
1534		else {
1535			np->type = NTO;
1536			pungetc();
1537		}
1538	} else {	/* c == '<' */
1539		np->nfile.fd = 0;
1540		c = pgetc();
1541		if (c == '<') {
1542			if (sizeof (struct nfile) != sizeof (struct nhere)) {
1543				np = (union node *)stalloc(sizeof (struct nhere));
1544				np->nfile.fd = 0;
1545			}
1546			np->type = NHERE;
1547			heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc));
1548			heredoc->here = np;
1549			if ((c = pgetc()) == '-') {
1550				heredoc->striptabs = 1;
1551			} else {
1552				heredoc->striptabs = 0;
1553				pungetc();
1554			}
1555		} else if (c == '&')
1556			np->type = NFROMFD;
1557		else if (c == '>')
1558			np->type = NFROMTO;
1559		else {
1560			np->type = NFROM;
1561			pungetc();
1562		}
1563	}
1564	if (fd != '\0')
1565		np->nfile.fd = digit_val(fd);
1566	redirnode = np;
1567	goto parseredir_return;
1568}
1569
1570
1571/*
1572 * Parse a substitution.  At this point, we have read the dollar sign
1573 * and nothing else.
1574 */
1575
1576parsesub: {
1577	char buf[10];
1578	int subtype;
1579	int typeloc;
1580	int flags;
1581	char *p;
1582	static const char types[] = "}-+?=";
1583	int bracketed_name = 0; /* used to handle ${[0-9]*} variables */
1584	int linno;
1585	int length;
1586	int c1;
1587
1588	c = pgetc();
1589	if (c == '(') {	/* $(command) or $((arith)) */
1590		if (pgetc() == '(') {
1591			PARSEARITH();
1592		} else {
1593			pungetc();
1594			out = parsebackq(out, &bqlist, 0,
1595			    state[level].syntax == DQSYNTAX &&
1596			    (eofmark == NULL || newvarnest > 0),
1597			    state[level].syntax == DQSYNTAX ||
1598			    state[level].syntax == ARISYNTAX);
1599		}
1600	} else if (c == '{' || is_name(c) || is_special(c)) {
1601		USTPUTC(CTLVAR, out);
1602		typeloc = out - stackblock();
1603		USTPUTC(VSNORMAL, out);
1604		subtype = VSNORMAL;
1605		flags = 0;
1606		if (c == '{') {
1607			bracketed_name = 1;
1608			c = pgetc();
1609			subtype = 0;
1610		}
1611varname:
1612		if (!is_eof(c) && is_name(c)) {
1613			length = 0;
1614			do {
1615				STPUTC(c, out);
1616				c = pgetc();
1617				length++;
1618			} while (!is_eof(c) && is_in_name(c));
1619			if (length == 6 &&
1620			    strncmp(out - length, "LINENO", length) == 0) {
1621				/* Replace the variable name with the
1622				 * current line number. */
1623				linno = plinno;
1624				if (funclinno != 0)
1625					linno -= funclinno - 1;
1626				snprintf(buf, sizeof(buf), "%d", linno);
1627				STADJUST(-6, out);
1628				STPUTS(buf, out);
1629				flags |= VSLINENO;
1630			}
1631		} else if (is_digit(c)) {
1632			if (bracketed_name) {
1633				do {
1634					STPUTC(c, out);
1635					c = pgetc();
1636				} while (is_digit(c));
1637			} else {
1638				STPUTC(c, out);
1639				c = pgetc();
1640			}
1641		} else if (is_special(c)) {
1642			c1 = c;
1643			c = pgetc();
1644			if (subtype == 0 && c1 == '#') {
1645				subtype = VSLENGTH;
1646				if (strchr(types, c) == NULL && c != ':' &&
1647				    c != '#' && c != '%')
1648					goto varname;
1649				c1 = c;
1650				c = pgetc();
1651				if (c1 != '}' && c == '}') {
1652					pungetc();
1653					c = c1;
1654					goto varname;
1655				}
1656				pungetc();
1657				c = c1;
1658				c1 = '#';
1659				subtype = 0;
1660			}
1661			USTPUTC(c1, out);
1662		} else {
1663			subtype = VSERROR;
1664			if (c == '}')
1665				pungetc();
1666			else if (c == '\n' || c == PEOF)
1667				synerror("Unexpected end of line in substitution");
1668			else
1669				USTPUTC(c, out);
1670		}
1671		if (subtype == 0) {
1672			switch (c) {
1673			case ':':
1674				flags |= VSNUL;
1675				c = pgetc();
1676				/*FALLTHROUGH*/
1677			default:
1678				p = strchr(types, c);
1679				if (p == NULL) {
1680					if (c == '\n' || c == PEOF)
1681						synerror("Unexpected end of line in substitution");
1682					if (flags == VSNUL)
1683						STPUTC(':', out);
1684					STPUTC(c, out);
1685					subtype = VSERROR;
1686				} else
1687					subtype = p - types + VSNORMAL;
1688				break;
1689			case '%':
1690			case '#':
1691				{
1692					int cc = c;
1693					subtype = c == '#' ? VSTRIMLEFT :
1694							     VSTRIMRIGHT;
1695					c = pgetc();
1696					if (c == cc)
1697						subtype++;
1698					else
1699						pungetc();
1700					break;
1701				}
1702			}
1703		} else if (subtype != VSERROR) {
1704			if (subtype == VSLENGTH && c != '}')
1705				subtype = VSERROR;
1706			pungetc();
1707		}
1708		STPUTC('=', out);
1709		if (state[level].syntax == DQSYNTAX ||
1710		    state[level].syntax == ARISYNTAX)
1711			flags |= VSQUOTE;
1712		*(stackblock() + typeloc) = subtype | flags;
1713		if (subtype != VSNORMAL) {
1714			if (level + 1 >= maxnest) {
1715				maxnest *= 2;
1716				if (state == state_static) {
1717					state = parser_temp_alloc(
1718					    maxnest * sizeof(*state));
1719					memcpy(state, state_static,
1720					    MAXNEST_static * sizeof(*state));
1721				} else
1722					state = parser_temp_realloc(state,
1723					    maxnest * sizeof(*state));
1724			}
1725			level++;
1726			state[level].parenlevel = 0;
1727			if (subtype == VSMINUS || subtype == VSPLUS ||
1728			    subtype == VSQUESTION || subtype == VSASSIGN) {
1729				/*
1730				 * For operators that were in the Bourne shell,
1731				 * inherit the double-quote state.
1732				 */
1733				state[level].syntax = state[level - 1].syntax;
1734				state[level].category = TSTATE_VAR_OLD;
1735			} else {
1736				/*
1737				 * The other operators take a pattern,
1738				 * so go to BASESYNTAX.
1739				 * Also, ' and " are now special, even
1740				 * in here documents.
1741				 */
1742				state[level].syntax = BASESYNTAX;
1743				state[level].category = TSTATE_VAR_NEW;
1744				newvarnest++;
1745			}
1746		}
1747	} else if (c == '\'' && state[level].syntax == BASESYNTAX) {
1748		/* $'cstylequotes' */
1749		USTPUTC(CTLQUOTEMARK, out);
1750		state[level].syntax = SQSYNTAX;
1751		sqiscstyle = 1;
1752	} else {
1753		USTPUTC('$', out);
1754		pungetc();
1755	}
1756	goto parsesub_return;
1757}
1758
1759
1760/*
1761 * Parse an arithmetic expansion (indicate start of one and set state)
1762 */
1763parsearith: {
1764
1765	if (level + 1 >= maxnest) {
1766		maxnest *= 2;
1767		if (state == state_static) {
1768			state = parser_temp_alloc(
1769			    maxnest * sizeof(*state));
1770			memcpy(state, state_static,
1771			    MAXNEST_static * sizeof(*state));
1772		} else
1773			state = parser_temp_realloc(state,
1774			    maxnest * sizeof(*state));
1775	}
1776	level++;
1777	state[level].syntax = ARISYNTAX;
1778	state[level].parenlevel = 0;
1779	state[level].category = TSTATE_ARITH;
1780	USTPUTC(CTLARI, out);
1781	if (state[level - 1].syntax == DQSYNTAX)
1782		USTPUTC('"',out);
1783	else
1784		USTPUTC(' ',out);
1785	goto parsearith_return;
1786}
1787
1788} /* end of readtoken */
1789
1790
1791/*
1792 * Returns true if the text contains nothing to expand (no dollar signs
1793 * or backquotes).
1794 */
1795
1796static int
1797noexpand(char *text)
1798{
1799	char *p;
1800	char c;
1801
1802	p = text;
1803	while ((c = *p++) != '\0') {
1804		if ( c == CTLQUOTEMARK)
1805			continue;
1806		if (c == CTLESC)
1807			p++;
1808		else if (BASESYNTAX[(int)c] == CCTL)
1809			return 0;
1810	}
1811	return 1;
1812}
1813
1814
1815/*
1816 * Return true if the argument is a legal variable name (a letter or
1817 * underscore followed by zero or more letters, underscores, and digits).
1818 */
1819
1820int
1821goodname(const char *name)
1822{
1823	const char *p;
1824
1825	p = name;
1826	if (! is_name(*p))
1827		return 0;
1828	while (*++p) {
1829		if (! is_in_name(*p))
1830			return 0;
1831	}
1832	return 1;
1833}
1834
1835
1836int
1837isassignment(const char *p)
1838{
1839	if (!is_name(*p))
1840		return 0;
1841	p++;
1842	for (;;) {
1843		if (*p == '=')
1844			return 1;
1845		else if (!is_in_name(*p))
1846			return 0;
1847		p++;
1848	}
1849}
1850
1851
1852static void
1853consumetoken(int token)
1854{
1855	if (readtoken() != token)
1856		synexpect(token);
1857}
1858
1859
1860/*
1861 * Called when an unexpected token is read during the parse.  The argument
1862 * is the token that is expected, or -1 if more than one type of token can
1863 * occur at this point.
1864 */
1865
1866static void
1867synexpect(int token)
1868{
1869	char msg[64];
1870
1871	if (token >= 0) {
1872		fmtstr(msg, 64, "%s unexpected (expecting %s)",
1873			tokname[lasttoken], tokname[token]);
1874	} else {
1875		fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]);
1876	}
1877	synerror(msg);
1878}
1879
1880
1881static void
1882synerror(const char *msg)
1883{
1884	if (commandname)
1885		outfmt(out2, "%s: %d: ", commandname, startlinno);
1886	outfmt(out2, "Syntax error: %s\n", msg);
1887	error((char *)NULL);
1888}
1889
1890static void
1891setprompt(int which)
1892{
1893	whichprompt = which;
1894
1895#ifndef NO_HISTORY
1896	if (!el)
1897#endif
1898	{
1899		out2str(getprompt(NULL));
1900		flushout(out2);
1901	}
1902}
1903
1904/*
1905 * called by editline -- any expansions to the prompt
1906 *    should be added here.
1907 */
1908char *
1909getprompt(void *unused __unused)
1910{
1911	static char ps[PROMPTLEN];
1912	char *fmt;
1913	const char *pwd;
1914	int i, trim;
1915	static char internal_error[] = "??";
1916
1917	/*
1918	 * Select prompt format.
1919	 */
1920	switch (whichprompt) {
1921	case 0:
1922		fmt = nullstr;
1923		break;
1924	case 1:
1925		fmt = ps1val();
1926		break;
1927	case 2:
1928		fmt = ps2val();
1929		break;
1930	default:
1931		return internal_error;
1932	}
1933
1934	/*
1935	 * Format prompt string.
1936	 */
1937	for (i = 0; (i < 127) && (*fmt != '\0'); i++, fmt++)
1938		if (*fmt == '\\')
1939			switch (*++fmt) {
1940
1941				/*
1942				 * Hostname.
1943				 *
1944				 * \h specifies just the local hostname,
1945				 * \H specifies fully-qualified hostname.
1946				 */
1947			case 'h':
1948			case 'H':
1949				ps[i] = '\0';
1950				gethostname(&ps[i], PROMPTLEN - i);
1951				/* Skip to end of hostname. */
1952				trim = (*fmt == 'h') ? '.' : '\0';
1953				while ((ps[i+1] != '\0') && (ps[i+1] != trim))
1954					i++;
1955				break;
1956
1957				/*
1958				 * Working directory.
1959				 *
1960				 * \W specifies just the final component,
1961				 * \w specifies the entire path.
1962				 */
1963			case 'W':
1964			case 'w':
1965				pwd = lookupvar("PWD");
1966				if (pwd == NULL)
1967					pwd = "?";
1968				if (*fmt == 'W' &&
1969				    *pwd == '/' && pwd[1] != '\0')
1970					strlcpy(&ps[i], strrchr(pwd, '/') + 1,
1971					    PROMPTLEN - i);
1972				else
1973					strlcpy(&ps[i], pwd, PROMPTLEN - i);
1974				/* Skip to end of path. */
1975				while (ps[i + 1] != '\0')
1976					i++;
1977				break;
1978
1979				/*
1980				 * Superuser status.
1981				 *
1982				 * '$' for normal users, '#' for root.
1983				 */
1984			case '$':
1985				ps[i] = (geteuid() != 0) ? '$' : '#';
1986				break;
1987
1988				/*
1989				 * A literal \.
1990				 */
1991			case '\\':
1992				ps[i] = '\\';
1993				break;
1994
1995				/*
1996				 * Emit unrecognized formats verbatim.
1997				 */
1998			default:
1999				ps[i++] = '\\';
2000				ps[i] = *fmt;
2001				break;
2002			}
2003		else
2004			ps[i] = *fmt;
2005	ps[i] = '\0';
2006	return (ps);
2007}
2008
2009
2010const char *
2011expandstr(const char *ps)
2012{
2013	union node n;
2014	struct jmploc jmploc;
2015	struct jmploc *const savehandler = handler;
2016	const int saveprompt = doprompt;
2017	struct parsefile *const savetopfile = getcurrentfile();
2018	struct parser_temp *const saveparser_temp = parser_temp;
2019	const char *result = NULL;
2020
2021	if (!setjmp(jmploc.loc)) {
2022		handler = &jmploc;
2023		parser_temp = NULL;
2024		setinputstring(ps, 1);
2025		doprompt = 0;
2026		readtoken1(pgetc(), DQSYNTAX, "\n\n", 0);
2027		if (backquotelist != NULL)
2028			error("Command substitution not allowed here");
2029
2030		n.narg.type = NARG;
2031		n.narg.next = NULL;
2032		n.narg.text = wordtext;
2033		n.narg.backquote = backquotelist;
2034
2035		expandarg(&n, NULL, 0);
2036		result = stackblock();
2037		INTOFF;
2038	}
2039	handler = savehandler;
2040	doprompt = saveprompt;
2041	popfilesupto(savetopfile);
2042	if (parser_temp != saveparser_temp) {
2043		parser_temp_free_all();
2044		parser_temp = saveparser_temp;
2045	}
2046	if (result != NULL) {
2047		INTON;
2048	} else if (exception == EXINT)
2049		raise(SIGINT);
2050	return result;
2051}
2052